Super-resolution reconstruction of a digital elevation model based on a deep residual network

Abstract

The digital elevation model (DEM) is an important basic data tool applied in geoscience applications. Because of its high cost and long development cycle of enhancing hardware performance, designing the related models and algorithms to improve the resolution of DEM is of considerable significance. At present, there is little research on DEM super-resolution based on deep learning, and the results of the reconstructed DEMs obtained by existing methods are inaccurate. Therefore, deepening of the network layers is utilized to improve the accuracy of a reconstructed DEM. This paper designs a neutral network model with 30 convolutional layers to learn the feature mapping relationship between a low- and high-resolution DEM. To avoid the problem of network degradation caused by increasing the number of convolutional layers, residual learning is introduced to accelerate the convergence speed of the model, thereby preferably realizing the DEM super-resolution process. The results show that DEM super-resolution based on a deep residual network is better than that obtained using a neural network with fewer convolutional layers, and the reconstructed result of the DEM based on a deep residual network is remarkably improved in terms of the peak signal to noise ratio and visual effect.

Deep Learning-Based Super-Resolution of Digital Elevation Models in Data Poor Regions.

“…Earlier research on Super Resolution of images has used different algorithms and data from different locations and with different peak values for PSNR, which make it difficult to compare our results with earlier work (Argudo et al, 2018;Chen et al, 2016;Demiray et al, 2020;Jiao et al, 2020;Kubade et al, 2020Kubade et al, , 2021Shin & Spittle, 2019b;Wu & Ma, 2020). Furthermore, unlike computer vision Super-Resolution, where existing test datasets are available for a fair comparison between algorithms, in the geoscience community, especially for DEM data, there is no standard dataset for comparison of the improvement, making it impossible to compare the performance of our approach relative to others.…”

“…Testing the model at the same location as it is trained also does not provide full information on the applicability of such methods to improve the quality of the freely available global dataset. Some researchers also used Shuttle Radar Topography Mission (SRTM) data which is available freely for super-resolution applications (Jiao et al, 2020;Wu & Ma, 2020), but in these cases, super sampling was done at very low resolution, producing a final resolution of 30 by 30 meters, which is too low for most/ many geoscience modelling studies.…”

Impacts of Resampling and Downscaling Digital Elevation Model and Its Morphometric Factors: A Comparison of Hopfield Neural Network, Bilinear, Bicubic, and Kriging Interpolations

“…More recently, a number of sophisticated deep learning artificial intelligence models, such as deep residual networks [22], Recursive Sub-Pixel Convolutional Neural Networks [23], Laplacian of Gaussian Super-resolution [24], Reconstruction Network Combining Internal and External Learning [25], Super-Resolution with Generative Adversarial Network [26] have been proposed. These methods are mostly based on approaches proposed for image super-resolution [22][23][24][25][26]. The result shows that the DEM accuracy has improved regarding root mean square error (RMSE) and the closeness to the reference data [19,20].…”

Super-resolution reconstruction based on Gaussian transform and attention mechanism

“…Image super-resolution (SR) reconstruction is a basic task of image processing and is widely used in image compression ( Chen et al, 2021a ; Shi, Li & Jiahuan, 2022 ; Chen et al, 2021b ), medical imaging ( Jiao et al, 2020 ), and other fields. It is a research hotspot in the field of image processing.…”